Do R&D tax credits work? Evidence from an international panel of

advertisement
Do R&D tax credits work? Evidence
from an international panel of
countries 1979-94.
Work in Progress – April 1998
Nicholas Bloom
Rachel Griffith
John Van Reenen
Policy Issues
• Substantial annual cost of current R&D tax
measures:
– $1bn estimated annual revenue loss from R&D
tax credit in USA.
– $1bn estimated annual revenue loss from R&D
tax measures in EU.
– Japan and Australia also have tax credits.
• Other countries such as the UK and
Germany considering an R&D tax credit.
Economic Issues
• Early studies indicated that R&D was not
very sensitive to R&D tax incentives
• More recent studies indicate a much greater
response. However, one country studies:
– problems disentangling tax policy and macro
shocks
– within firm variation in after tax price
endogenous (?)
– generalisable to other countries (?)
What’s new in this paper?
• Extend analysis to more countries
• R&D tax changes as natural experiments
• Appropriate R&D data
– recorded on the basis of where it is conducted
– consistent definition across countries
– not tax data so not so susceptible to relabelling
• Business R&D data by source of funding
Tax Data
• New data set developed at the IFS which constructs a “user
cost of R&D”, reflecting
– general features of the tax system
– treatment of cross border flows
– special treatment of R&D investment including tax
credits and accelerated depreciation allowances
• Eight countries studied over 16 year, 1979-1994.
– Five with some kind of R&D tax promotion: Australia,
Canada, France, Japan, and the USA
– Three without any major R&D tax promotion: Italy,
Germany, and the UK
The User Cost of R&D
(without inflation or exchange rates)
 
1 A
1
(r   )
where:
A = Net present value of depreciation allowances and tax credits
 = Statutory tax rate
r = Real interest rate
 = Economic depreciation rate
User cost of R&D
.4
.35
.3
.25
.2
.15
1980
1985
1990
year
1995
Tax variation in user cost
1
.8
.6
.4
1980
1985
1990
year
1995
R&D data
• OECD data on R&D.
– business conducted R&D (BERD)
– manufacturing sector
– government survey data based on common
Frascati definition.
– deflated by 50% wages and 50% GDP deflator.
Manufacturing BERD/
Manufacturing output
4
3.5
USA
Japan
Germany
France
UK
3
2.5
2
Canada
Australia
Italy
1.5
1
.5
0
1980
1985
1990
year
1994
R&D data
• OECD MSTI data on the funding
breakdown of business conducted R&D
(BERD) into:
– domestic business financed BERD
– foreign business financed BERD
– government financed BERD
Total R&D Breakdown
(UK 1994, £14Bn total)
Domestic
Business
Financed
47%
GOVERD
15%
BERD
66%
HERD
19%
Government
Financed
11%
Foreign
Business
Financed
8%
Table 1 from paper
The model
ritd    ritd1  yit   itd  fi  t t   it
business funded lag
business R&D
user cost
of R&D
output
country
specific
effect
time
specific
effect
Relocation?
• Is the impact of R&D tax credit only to
change level of domestic R&D?
• Is there also an impact through relocating
R&D done by multinational firms?
Adding in foreign tax price
ritd    ritd1  yit 1   itd   itf  fi  tt   it
domestic
user cost
 0
foreign
user cost
indicates in favour of relocation
Measuring foreign tax price
7
   FDIij  ijt
f
it
j 1
Foreign Direct
Investment from
country i to
country j
User cost of R&D
for investment from
country i to country j
Table 2 from paper
Conclusions
• Using tax changes as exogenous changes to
the user cost of R&D we find an impact
elasticity of around -0.1 and long-run
elasticity or around -1.0
• Evidence suggests that the location of R&D
may be affected by tax-induced changes to
the cost of R&D, maybe some tax rivalry
Download